Assessing the efficiency of shape-based functions and descriptors in multi-scale matching of linear objects

The study of objects similarity in terms of shape has various applications such as determining the similarity degree in object matching. To this end, different functions and descriptors have been used. However, efficiency of each method used in various studies for solving linear object matching in datasets with either different or similar scales or sources has not been studied yet. This article studies the efficiency of the most important functions (i.e. turning, signature, and tangent) along with shape descriptors (i.e. shape context, LORD, and shape signature) in vector datasets with different scales and sources. For this purpose, three datasets of roads network with different sources and scales were employed. Results showed the greater efficiency of the turning function compared to the other methods. In addition to being able to identify corresponding objects in multi-scale datasets, it can improve matching and is capable of discovering shape difference in non-corresponding objects.